In [1]:
  !pip install -q -U keras-tuner
In [58]:
pip install tflearn

Requirement already satisfied: tflearn in c:\users\ashwinishivaprasad\anaconda\lib\site-packages (0.5.0)
Requirement already satisfied: numpy in c:\users\ashwinishivaprasad\anaconda\lib\site-packages (from tflearn) (1.19.5)
Requirement already satisfied: six in c:\users\ashwinishivaprasad\anaconda\lib\site-packages (from tflearn) (1.15.0)
Requirement already satisfied: Pillow in c:\users\ashwinishivaprasad\anaconda\lib\site-packages (from tflearn) (7.0.0)
Note: you may need to restart the kernel to use updated packages.
In [54]:
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
%matplotlib inline
import seaborn as sns
from sklearn.linear_model import  LogisticRegression,LinearRegression
from sklearn.svm import  SVC
from sklearn.tree import  DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import  Dense, BatchNormalization,Dropout, Convolution2D,Input,Conv2D,MaxPooling2D,Flatten
from tensorflow.keras.optimizers import  Adam, SGD
from tensorflow.keras.initializers import GlorotUniform, GlorotNormal,HeNormal,HeUniform
from tensorflow.keras.regularizers import L2
from sklearn.preprocessing import StandardScaler
from sklearn.decomposition import PCA
from sklearn.model_selection import  train_test_split,GridSearchCV,cross_val_score
import tensorflow
import cv2
from tensorflow.keras.preprocessing.image import ImageDataGenerator
from keras.applications import VGG19
from sklearn.metrics import accuracy_score
from sklearn.neighbors import KNeighborsClassifier
import tkinter as tk
from tkinter import ttk,Tk
from google.colab.patches import cv2_imshow
from google.colab import files
from tensorflow.keras.models import load_model
from PIL import Image
import kerastuner as kt
from kerastuner import RandomSearch
from kerastuner.engine.hyperparameters import HyperParameters
In [4]:
from google.colab import drive 
drive.mount('/content/drive',force_remount=True)

Mounted at /content/drive

1. Import the data and analyse the dimension of data.

In [60]:
#Import dataset 
import tflearn.datasets.oxflower17 as oxflower17

WARNING:tensorflow:From C:\Users\AshwiniShivaprasad\anaconda\lib\site-packages\tensorflow\python\compat\v2_compat.py:96: disable_resource_variables (from tensorflow.python.ops.variable_scope) is deprecated and will be removed in a future version.
Instructions for updating:
non-resource variables are not supported in the long term
curses is not supported on this machine (please install/reinstall curses for an optimal experience)
In [61]:
(X,Y) = oxflower17.load_data()

Downloading Oxford 17 category Flower Dataset, Please wait...

100.0% 60276736 / 60270631

Succesfully downloaded 17flowers.tgz 60270631 bytes.
File Extracted
Starting to parse images...
Parsing Done!
In [7]:
#Dimension of data
print("Dimensions of data :", X.shape)
print("Dimensions of label :", Y.shape)

Dimensions of data : (1360, 224, 224, 3)
Dimensions of label : (1360,)
In [ ]:
# The data contains 1360 rows of images of size 224*224*3 pixels, 224*224 being the size of each image and 3 being the channel size.

2. Preprocess the data.

In [62]:
trainX,testX,trainY,testY = train_test_split(X,Y, test_size=0.25)
In [10]:
print("Minimum value in training set :",trainX.min())
print("Maximum value in training set :",trainX.max())
print("Minimum value in test set :",testX.min())
print("Maximum value in test set :",testX.max())

Minimum value in training set : 0.0
Maximum value in training set : 1.0
Minimum value in test set : 0.0
Maximum value in test set : 1.0
Since min and max are already in range 0-1 , we do not need to normalize the data.
In [11]:
#Print shapes post splitting
print("X train shape :" ,trainX.shape)
print("Y train shape :" ,trainY.shape)
print("X test  shape :" ,testX.shape)
print("Y test shape :" ,testY.shape)

X train shape : (1020, 224, 224, 3)
Y train shape : (1020,)
X test  shape : (340, 224, 224, 3)
Y test shape : (340,)
In [12]:
#print first 5  values in test target set
testY[:5]
Out[12]:
array([15,  9,  0,  3,  9])
In [13]:
pd.value_counts(trainY)
Out[13]:
14    67
11    66
2     65
13    64
4     64
8     63
1     62
10    61
12    60
5     60
7     59
3     58
9     58
0     56
16    55
6     51
15    51
dtype: int64
In [14]:
pd.value_counts(testY)
Out[14]:
15    29
6     29
16    25
0     24
9     22
3     22
7     21
12    20
5     20
10    19
1     18
8     17
13    16
4     16
2     15
11    14
14    13
dtype: int64

3. Data visualisation

In [15]:
#Display Images
for i in range(400,410):

  plt.imshow(trainX[i],cmap='gray')
  plt.show()
In [17]:
#Display labels

for i in range(400,410):
  print("Label: ", trainY[i])

Label:  15
Label:  3
Label:  16
Label:  12
Label:  7
Label:  5
Label:  9
Label:  2
Label:  8
Label:  16
In [18]:
#Images vs Labels
for i in range(200,210):
  print("Label :",trainY[i])
  plt.imshow(trainX[i],cmap='gray')
  plt.show()

Label : 5

Label : 6

Label : 5

Label : 15

Label : 3

Label : 13

Label : 9

Label : 0

Label : 6

Label : 0
In [19]:
#Apply different filters on images and display the images
#Blur
flowerImg = trainX[100]
blurredImg = cv2.blur(flowerImg,(9,9))
plt.subplot(121),plt.imshow(flowerImg),plt.title('Original')
plt.xticks([]), plt.yticks([])
plt.subplot(122),plt.imshow(blurredImg),plt.title('Blurred')
plt.xticks([]), plt.yticks([])
plt.show()
In [63]:
#Edge Detection: Vertical
edge_matrix=  np.array([[-1, 0, 1],
                   [-1, 0, 1],
                   [-1, 0, 1]])
flowerImg2= trainX[200]
verticalEdge = cv2.filter2D(flowerImg2, -1, edge_matrix)
plt.subplot(121),plt.imshow(flowerImg2),plt.title('Original')
plt.xticks([]), plt.yticks([])
plt.subplot(122),plt.imshow(verticalEdge),plt.title('Picture with vertical edges')
plt.xticks([]), plt.yticks([])
plt.show()

Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
In [64]:
#Edge Detection: Horizantal
edge_matrix_hor=  np.array([[-1,-1, 1],
                   [0, 0, 0],
                   [-1,1, 1]])
HorizaontalEdge = cv2.filter2D(flowerImg2, -1, edge_matrix_hor)
plt.subplot(121),plt.imshow(flowerImg2),plt.title('Original')
plt.xticks([]), plt.yticks([])
plt.subplot(122),plt.imshow(HorizaontalEdge),plt.title('Picture with horizontal edges')
plt.xticks([]), plt.yticks([])
plt.show()

Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
In [65]:
#Emboss
emboss_matrix=  np.array([[-1,-1,0, 0],
                   [1,1,1,1],
                   [1,2,2,2],
                   [0,0,1,1]])
embossImg = cv2.filter2D(flowerImg2, -1, emboss_matrix)
plt.subplot(121),plt.imshow(flowerImg2),plt.title('Original')
plt.xticks([]), plt.yticks([])
plt.subplot(122),plt.imshow(embossImg),plt.title('Embossed picture')
plt.xticks([]), plt.yticks([])
plt.show()

Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).
In [66]:
#Smoothening

smooth_matrix = np.ones((5,5),np.float32)/25
smoothImg = cv2.filter2D(flowerImg2,-1,smooth_matrix)

plt.subplot(121),plt.imshow(flowerImg2),plt.title('Original')
plt.xticks([]), plt.yticks([])
plt.subplot(122),plt.imshow(smoothImg),plt.title('Averaging')
plt.xticks([]), plt.yticks([])
plt.show()

4. Train tune and test AIML image classifier model using:

• Supervised learning algorithms for training
• Use neural networks for training
• Use CNN for training
• Use various CNN with transferred learning models for training

4.1 Supervised learning algorithm

In [25]:
X_shaped=X.reshape(1360,150528)
In [26]:
#Standdard scaling and dimensionality reduction
scale = StandardScaler()
flowersScale = scale.fit_transform(X_shaped)
pca = PCA(n_components=50)
flowers_pca = pca.fit_transform(flowersScale)
print('PCA  shape for train ds is: ', flowers_pca.shape)
(trainX_scale, testX_scale, trainY_scale, testY_scale) = train_test_split(flowers_pca, Y, test_size=0.25) 

PCA  shape for train ds is:  (1360, 50)
In [27]:
#SVC classifier model
linearsvc=SVC(C=0.90,decision_function_shape='ovo',kernel='rbf')
linearsvc.fit(trainX_scale,trainY_scale)
svc_pred=linearsvc.predict(testX_scale)
svc_pred_train=linearsvc.predict(trainX_scale)
print('SVC train Accuracy: %.3f' % accuracy_score(trainY_scale,svc_pred_train))
print('SVC test Accuracy: %.3f' % accuracy_score(testY_scale,svc_pred))

SVC train Accuracy: 0.790
SVC test Accuracy: 0.450
In [28]:
from sklearn.metrics import classification_report
print("Classification report : \n",classification_report(testY_scale,svc_pred))

Classification report : 
               precision    recall  f1-score   support

           0       0.36      0.38      0.37        21
           1       0.03      0.07      0.05        14
           2       0.70      0.64      0.67        22
           3       0.39      0.44      0.41        16
           4       0.33      0.59      0.43        17
           5       0.22      0.27      0.24        22
           6       0.35      0.33      0.34        18
           7       0.67      0.67      0.67        18
           8       0.71      0.59      0.65        17
           9       0.45      0.56      0.50        16
          10       0.32      0.56      0.41        16
          11       0.50      0.30      0.38        23
          12       0.77      0.68      0.72        25
          13       0.60      0.55      0.57        22
          14       0.29      0.07      0.12        27
          15       0.41      0.33      0.37        21
          16       1.00      0.64      0.78        25

    accuracy                           0.45       340
   macro avg       0.48      0.45      0.45       340
weighted avg       0.50      0.45      0.46       340
In [ ]:
#SVC with Hyperparameter tuning
In [29]:
tuned_parameters = [{'kernel': ['rbf'],'C': [0.1,0.5,0.8,0.9,1, 10, 100, 1000]}]
In [30]:
clf = GridSearchCV(SVC(), tuned_parameters, verbose = 3)
clf.fit(trainX_scale,trainY_scale)

Fitting 5 folds for each of 8 candidates, totalling 40 fits
[CV] C=0.1, kernel=rbf ...............................................
[CV] ................... C=0.1, kernel=rbf, score=0.123, total=   0.2s
[CV] C=0.1, kernel=rbf ...............................................

[Parallel(n_jobs=1)]: Using backend SequentialBackend with 1 concurrent workers.
[Parallel(n_jobs=1)]: Done   1 out of   1 | elapsed:    0.2s remaining:    0.0s

[CV] ................... C=0.1, kernel=rbf, score=0.123, total=   0.2s
[CV] C=0.1, kernel=rbf ...............................................
[CV] ................... C=0.1, kernel=rbf, score=0.103, total=   0.2s
[CV] C=0.1, kernel=rbf ...............................................

[Parallel(n_jobs=1)]: Done   2 out of   2 | elapsed:    0.3s remaining:    0.0s

[CV] ................... C=0.1, kernel=rbf, score=0.083, total=   0.2s
[CV] C=0.1, kernel=rbf ...............................................
[CV] ................... C=0.1, kernel=rbf, score=0.083, total=   0.2s
[CV] C=0.5, kernel=rbf ...............................................
[CV] ................... C=0.5, kernel=rbf, score=0.441, total=   0.2s
[CV] C=0.5, kernel=rbf ...............................................
[CV] ................... C=0.5, kernel=rbf, score=0.446, total=   0.1s
[CV] C=0.5, kernel=rbf ...............................................
[CV] ................... C=0.5, kernel=rbf, score=0.451, total=   0.1s
[CV] C=0.5, kernel=rbf ...............................................
[CV] ................... C=0.5, kernel=rbf, score=0.417, total=   0.2s
[CV] C=0.5, kernel=rbf ...............................................
[CV] ................... C=0.5, kernel=rbf, score=0.436, total=   0.1s
[CV] C=0.8, kernel=rbf ...............................................
[CV] ................... C=0.8, kernel=rbf, score=0.461, total=   0.1s
[CV] C=0.8, kernel=rbf ...............................................
[CV] ................... C=0.8, kernel=rbf, score=0.475, total=   0.1s
[CV] C=0.8, kernel=rbf ...............................................
[CV] ................... C=0.8, kernel=rbf, score=0.510, total=   0.1s
[CV] C=0.8, kernel=rbf ...............................................
[CV] ................... C=0.8, kernel=rbf, score=0.441, total=   0.1s
[CV] C=0.8, kernel=rbf ...............................................
[CV] ................... C=0.8, kernel=rbf, score=0.456, total=   0.1s
[CV] C=0.9, kernel=rbf ...............................................
[CV] ................... C=0.9, kernel=rbf, score=0.461, total=   0.2s
[CV] C=0.9, kernel=rbf ...............................................
[CV] ................... C=0.9, kernel=rbf, score=0.471, total=   0.1s
[CV] C=0.9, kernel=rbf ...............................................
[CV] ................... C=0.9, kernel=rbf, score=0.529, total=   0.2s
[CV] C=0.9, kernel=rbf ...............................................
[CV] ................... C=0.9, kernel=rbf, score=0.456, total=   0.1s
[CV] C=0.9, kernel=rbf ...............................................
[CV] ................... C=0.9, kernel=rbf, score=0.456, total=   0.1s
[CV] C=1, kernel=rbf .................................................
[CV] ..................... C=1, kernel=rbf, score=0.461, total=   0.2s
[CV] C=1, kernel=rbf .................................................
[CV] ..................... C=1, kernel=rbf, score=0.480, total=   0.1s
[CV] C=1, kernel=rbf .................................................
[CV] ..................... C=1, kernel=rbf, score=0.534, total=   0.2s
[CV] C=1, kernel=rbf .................................................
[CV] ..................... C=1, kernel=rbf, score=0.471, total=   0.1s
[CV] C=1, kernel=rbf .................................................
[CV] ..................... C=1, kernel=rbf, score=0.441, total=   0.1s
[CV] C=10, kernel=rbf ................................................
[CV] .................... C=10, kernel=rbf, score=0.559, total=   0.2s
[CV] C=10, kernel=rbf ................................................
[CV] .................... C=10, kernel=rbf, score=0.495, total=   0.2s
[CV] C=10, kernel=rbf ................................................
[CV] .................... C=10, kernel=rbf, score=0.544, total=   0.2s
[CV] C=10, kernel=rbf ................................................
[CV] .................... C=10, kernel=rbf, score=0.475, total=   0.2s
[CV] C=10, kernel=rbf ................................................
[CV] .................... C=10, kernel=rbf, score=0.471, total=   0.2s
[CV] C=100, kernel=rbf ...............................................
[CV] ................... C=100, kernel=rbf, score=0.549, total=   0.2s
[CV] C=100, kernel=rbf ...............................................
[CV] ................... C=100, kernel=rbf, score=0.495, total=   0.2s
[CV] C=100, kernel=rbf ...............................................
[CV] ................... C=100, kernel=rbf, score=0.544, total=   0.2s
[CV] C=100, kernel=rbf ...............................................
[CV] ................... C=100, kernel=rbf, score=0.475, total=   0.2s
[CV] C=100, kernel=rbf ...............................................
[CV] ................... C=100, kernel=rbf, score=0.466, total=   0.2s
[CV] C=1000, kernel=rbf ..............................................
[CV] .................. C=1000, kernel=rbf, score=0.549, total=   0.2s
[CV] C=1000, kernel=rbf ..............................................
[CV] .................. C=1000, kernel=rbf, score=0.495, total=   0.2s
[CV] C=1000, kernel=rbf ..............................................
[CV] .................. C=1000, kernel=rbf, score=0.544, total=   0.2s
[CV] C=1000, kernel=rbf ..............................................
[CV] .................. C=1000, kernel=rbf, score=0.475, total=   0.2s
[CV] C=1000, kernel=rbf ..............................................
[CV] .................. C=1000, kernel=rbf, score=0.466, total=   0.2s

[Parallel(n_jobs=1)]: Done  40 out of  40 | elapsed:    6.4s finished
Out[30]:
GridSearchCV(cv=None, error_score=nan,
             estimator=SVC(C=1.0, break_ties=False, cache_size=200,
                           class_weight=None, coef0=0.0,
                           decision_function_shape='ovr', degree=3,
                           gamma='scale', kernel='rbf', max_iter=-1,
                           probability=False, random_state=None, shrinking=True,
                           tol=0.001, verbose=False),
             iid='deprecated', n_jobs=None,
             param_grid=[{'C': [0.1, 0.5, 0.8, 0.9, 1, 10, 100, 1000],
                          'kernel': ['rbf']}],
             pre_dispatch='2*n_jobs', refit=True, return_train_score=False,
             scoring=None, verbose=3)
In [31]:
clf.best_score_
Out[31]:
0.5088235294117648
In [32]:
clf.best_params_
Out[32]:
{'C': 10, 'kernel': 'rbf'}
In [33]:
results= clf.cv_results_
results
Out[33]:
{'mean_fit_time': array([0.148105  , 0.12839942, 0.12510471, 0.12960567, 0.12837868,
        0.15350866, 0.15217443, 0.1523303 ]),
 'mean_score_time': array([0.02130175, 0.01987605, 0.02021055, 0.01971731, 0.02118855,
        0.02045035, 0.0193893 , 0.01988358]),
 'mean_test_score': array([0.10294118, 0.43823529, 0.46862745, 0.4745098 , 0.47745098,
        0.50882353, 0.50588235, 0.50588235]),
 'param_C': masked_array(data=[0.1, 0.5, 0.8, 0.9, 1, 10, 100, 1000],
              mask=[False, False, False, False, False, False, False, False],
        fill_value='?',
             dtype=object),
 'param_kernel': masked_array(data=['rbf', 'rbf', 'rbf', 'rbf', 'rbf', 'rbf', 'rbf', 'rbf'],
              mask=[False, False, False, False, False, False, False, False],
        fill_value='?',
             dtype=object),
 'params': [{'C': 0.1, 'kernel': 'rbf'},
  {'C': 0.5, 'kernel': 'rbf'},
  {'C': 0.8, 'kernel': 'rbf'},
  {'C': 0.9, 'kernel': 'rbf'},
  {'C': 1, 'kernel': 'rbf'},
  {'C': 10, 'kernel': 'rbf'},
  {'C': 100, 'kernel': 'rbf'},
  {'C': 1000, 'kernel': 'rbf'}],
 'rank_test_score': array([8, 7, 6, 5, 4, 1, 2, 2], dtype=int32),
 'split0_test_score': array([0.12254902, 0.44117647, 0.46078431, 0.46078431, 0.46078431,
        0.55882353, 0.54901961, 0.54901961]),
 'split1_test_score': array([0.12254902, 0.44607843, 0.4754902 , 0.47058824, 0.48039216,
        0.49509804, 0.49509804, 0.49509804]),
 'split2_test_score': array([0.10294118, 0.45098039, 0.50980392, 0.52941176, 0.53431373,
        0.54411765, 0.54411765, 0.54411765]),
 'split3_test_score': array([0.08333333, 0.41666667, 0.44117647, 0.45588235, 0.47058824,
        0.4754902 , 0.4754902 , 0.4754902 ]),
 'split4_test_score': array([0.08333333, 0.43627451, 0.45588235, 0.45588235, 0.44117647,
        0.47058824, 0.46568627, 0.46568627]),
 'std_fit_time': array([0.00286575, 0.00240919, 0.0016705 , 0.00412595, 0.00178328,
        0.0029788 , 0.00037907, 0.00353093]),
 'std_score_time': array([0.00155851, 0.000739  , 0.00098972, 0.00021114, 0.00237406,
        0.00224125, 0.00023685, 0.00067711]),
 'std_test_score': array([0.01753779, 0.01184612, 0.02332427, 0.02797126, 0.03124976,
        0.03607523, 0.03457881, 0.03457881])}
In [34]:
linearsvc=SVC(C=10,decision_function_shape='ovo',kernel='rbf')
scores=cross_val_score(linearsvc,trainX_scale,trainY_scale,cv=15)
scores
Out[34]:
array([0.57352941, 0.57352941, 0.54411765, 0.44117647, 0.51470588,
       0.48529412, 0.57352941, 0.54411765, 0.51470588, 0.51470588,
       0.52941176, 0.51470588, 0.48529412, 0.44117647, 0.39705882])
In [35]:
np.mean(scores)
Out[35]:
0.5098039215686275

4.2 Neural Network

In [16]:
#Neural network
trainX_reshape =trainX.reshape(1020,150528)    #224*224*3 =150528
testX_reshaped = testX.reshape(340,150528)
In [17]:
trainY_encode=tensorflow.keras.utils.to_categorical(trainY,num_classes=17)
testY_encode =tensorflow.keras.utils.to_categorical(testY,num_classes=17)
In [69]:
def build_NNmodel(hp):  
  model = Sequential([
    
    Dense(
        units=hp.Int('Dense1', min_value=32, max_value=512, step=32),
        activation='relu',
        input_shape =(150528,),
        kernel_initializer =hp.Choice('wt_initialiser', values=['glorot_uniform','glorot_normal','he_uniform','he_normal']),
       # kernel_regularizer=L2(hp.Choice('regulariser', values=[1E-2,2e-2, 1E-3,2e-3 ,1E-4, 2e-4,1E-5,2e-5]))
    ),
    # BatchNormalization(),
   # Dropout(hp.Choice('dropoutRate1', values=[0.4,.5,.6])),
     Dense(
        units=hp.Int('Dense2', min_value=32, max_value=512, step=32),
        activation='relu',
        kernel_initializer =hp.Choice('wt_initialiser1', values=['glorot_uniform','glorot_normal','he_uniform','he_normal']),
       # kernel_regularizer=L2(hp.Choice('regulariser1', values=[1E-2,2e-2, 1E-3,2e-3 ,1E-4, 2e-4,1E-5,2e-5]))
        ),
#		BatchNormalization(),
 #   Dropout(hp.Choice('dropoutRate2', values=[0.4,.5,.6])),
     Dense(
        units=hp.Int('Dense3', min_value=32, max_value=512, step=32),
        activation='relu',
        kernel_initializer =hp.Choice('wt_initialiser2', values=['glorot_uniform','glorot_normal','he_uniform','he_normal']),
        #kernel_regularizer=L2(hp.Choice('regulariser2', values=[1E-2, 1E-3,1E-4, 1E-5]))
    ),
   # BatchNormalization(),
  #  Dropout(hp.Choice('dropoutRate3', values=[0.4,.5,.6])),
     Dense(
        units=hp.Int('Dense4', min_value=32, max_value=512, step=32),
        activation='relu',    
        kernel_initializer =hp.Choice('wt_initialiser3', values=['glorot_uniform','glorot_normal','he_uniform','he_normal'])   ,
        #kernel_regularizer=L2(hp.Choice('regulariser3', values=[1E-2,1E-3,1E-4, 1E-5]))     
    ),
 #  BatchNormalization(),
 # Dropout(hp.Choice('dropoutRate4', values=[0.4,.5,.6])),
 

   # BatchNormalization(),
    Dense(17, activation='softmax',kernel_regularizer=L2(hp.Choice('regulariser8', values=[1E-2, 1E-3,1E-4, 1E-5])),
          kernel_initializer =hp.Choice('wt_initialiser8', values=['glorot_uniform','glorot_normal','he_uniform','he_normal']))
  ])
  
  model.compile(optimizer=SGD(hp.Choice('learning_rate',  values=[1E-2,1E-3,1E-4])),
                          loss='categorical_crossentropy',
                           metrics=['accuracy'])
  
  return model
In [70]:
tuner_search=RandomSearch(build_NNmodel,objective='acc',max_trials=5,directory='output65',project_name="Flower")
In [71]:
    tuner_search.search(trainX_reshape,trainY_encode,epochs=20,validation_data=(testX_reshaped,testY_encode))

Trial 5 Complete [00h 02m 32s]
acc: 0.6647058725357056

Best acc So Far: 0.6647058725357056
Total elapsed time: 00h 08m 57s
INFO:tensorflow:Oracle triggered exit
In [72]:
modelFlowerNN = tuner_search.get_best_models(num_models=1)[0]
In [75]:
modelFlowerNN.fit(trainX_reshape,trainY_encode,epochs=50,validation_data=(testX_reshaped,testY_encode))

Train on 1020 samples, validate on 340 samples
Epoch 1/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.1977 - acc: 0.9598 - val_loss: 2.2998 - val_acc: 0.4824
Epoch 2/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.2092 - acc: 0.9559 - val_loss: 1.8488 - val_acc: 0.5176
Epoch 3/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.1382 - acc: 0.9765 - val_loss: 1.8594 - val_acc: 0.5382
Epoch 4/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.1200 - acc: 0.9824 - val_loss: 1.8668 - val_acc: 0.5324
Epoch 5/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0930 - acc: 0.9882 - val_loss: 1.8846 - val_acc: 0.5441
Epoch 6/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0903 - acc: 0.9931 - val_loss: 1.8654 - val_acc: 0.5382
Epoch 7/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0636 - acc: 0.9980 - val_loss: 2.1174 - val_acc: 0.5294
Epoch 8/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0597 - acc: 0.9990 - val_loss: 1.9448 - val_acc: 0.5500
Epoch 9/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0571 - acc: 0.9980 - val_loss: 2.1162 - val_acc: 0.5412
Epoch 10/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0551 - acc: 1.0000 - val_loss: 2.0202 - val_acc: 0.5382
Epoch 11/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0511 - acc: 1.0000 - val_loss: 2.0124 - val_acc: 0.5412
Epoch 12/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0494 - acc: 1.0000 - val_loss: 2.0136 - val_acc: 0.5529
Epoch 13/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0481 - acc: 1.0000 - val_loss: 2.0434 - val_acc: 0.5441
Epoch 14/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0471 - acc: 1.0000 - val_loss: 2.0676 - val_acc: 0.5500
Epoch 15/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0463 - acc: 1.0000 - val_loss: 2.0623 - val_acc: 0.5441
Epoch 16/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0453 - acc: 1.0000 - val_loss: 2.0749 - val_acc: 0.5441
Epoch 17/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0445 - acc: 1.0000 - val_loss: 2.0878 - val_acc: 0.5412
Epoch 18/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0440 - acc: 1.0000 - val_loss: 2.0999 - val_acc: 0.5412
Epoch 19/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0435 - acc: 1.0000 - val_loss: 2.1229 - val_acc: 0.5412
Epoch 20/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0431 - acc: 1.0000 - val_loss: 2.1194 - val_acc: 0.5412
Epoch 21/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0425 - acc: 1.0000 - val_loss: 2.1260 - val_acc: 0.5412
Epoch 22/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0422 - acc: 1.0000 - val_loss: 2.1483 - val_acc: 0.5412
Epoch 23/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0418 - acc: 1.0000 - val_loss: 2.1489 - val_acc: 0.5441
Epoch 24/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0415 - acc: 1.0000 - val_loss: 2.1396 - val_acc: 0.5471
Epoch 25/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0412 - acc: 1.0000 - val_loss: 2.1404 - val_acc: 0.5382
Epoch 26/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0409 - acc: 1.0000 - val_loss: 2.1522 - val_acc: 0.5412
Epoch 27/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0407 - acc: 1.0000 - val_loss: 2.1756 - val_acc: 0.5412
Epoch 28/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0405 - acc: 1.0000 - val_loss: 2.1627 - val_acc: 0.5382
Epoch 29/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0402 - acc: 1.0000 - val_loss: 2.1777 - val_acc: 0.5412
Epoch 30/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0400 - acc: 1.0000 - val_loss: 2.1781 - val_acc: 0.5382
Epoch 31/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0398 - acc: 1.0000 - val_loss: 2.1928 - val_acc: 0.5412
Epoch 32/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0395 - acc: 1.0000 - val_loss: 2.1939 - val_acc: 0.5353
Epoch 33/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0394 - acc: 1.0000 - val_loss: 2.2057 - val_acc: 0.5412
Epoch 34/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0392 - acc: 1.0000 - val_loss: 2.2061 - val_acc: 0.5353
Epoch 35/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0391 - acc: 1.0000 - val_loss: 2.2076 - val_acc: 0.5412
Epoch 36/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0389 - acc: 1.0000 - val_loss: 2.2155 - val_acc: 0.5412
Epoch 37/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0388 - acc: 1.0000 - val_loss: 2.2233 - val_acc: 0.5353
Epoch 38/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0385 - acc: 1.0000 - val_loss: 2.2236 - val_acc: 0.5324
Epoch 39/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0385 - acc: 1.0000 - val_loss: 2.2336 - val_acc: 0.5265
Epoch 40/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0383 - acc: 1.0000 - val_loss: 2.2360 - val_acc: 0.5382
Epoch 41/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0382 - acc: 1.0000 - val_loss: 2.2409 - val_acc: 0.5412
Epoch 42/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0381 - acc: 1.0000 - val_loss: 2.2455 - val_acc: 0.5412
Epoch 43/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0380 - acc: 1.0000 - val_loss: 2.2461 - val_acc: 0.5382
Epoch 44/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0379 - acc: 1.0000 - val_loss: 2.2626 - val_acc: 0.5382
Epoch 45/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0377 - acc: 1.0000 - val_loss: 2.2519 - val_acc: 0.5353
Epoch 46/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0377 - acc: 1.0000 - val_loss: 2.2584 - val_acc: 0.5382
Epoch 47/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0376 - acc: 1.0000 - val_loss: 2.2645 - val_acc: 0.5324
Epoch 48/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0374 - acc: 1.0000 - val_loss: 2.2648 - val_acc: 0.5412
Epoch 49/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0374 - acc: 1.0000 - val_loss: 2.2671 - val_acc: 0.5382
Epoch 50/50
1020/1020 [==============================] - 7s 7ms/sample - loss: 0.0372 - acc: 1.0000 - val_loss: 2.2718 - val_acc: 0.5382
Out[75]:
<tensorflow.python.keras.callbacks.History at 0x7fdd0c023b90>
In [76]:
print("Test set accuracy :",modelFlowerNN.evaluate(testX_reshaped,testY_encode))

Test set accuracy : [2.2717643653645236, 0.5382353]
In [77]:
prediction = np.argmax(modelFlowerNN.predict(testX_reshaped), axis=-1)

/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/training.py:2325: UserWarning: `Model.state_updates` will be removed in a future version. This property should not be used in TensorFlow 2.0, as `updates` are applied automatically.
  warnings.warn('`Model.state_updates` will be removed in a future version. '
In [78]:
for i in range(5,10):
  print("Actual Label :",testY_encode[i])
  print("Predicted label: ",prediction[i])

Actual Label : [0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 0. 0. 0.]
Predicted label:  12
Actual Label : [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 0. 0. 0.]
Predicted label:  14
Actual Label : [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 0. 0.]
Predicted label:  11
Actual Label : [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 0.]
Predicted label:  7
Actual Label : [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 0.]
Predicted label:  13

4.3 CNN

In [72]:
train_gen=ImageDataGenerator(
        rotation_range=40,
        width_shift_range=0.2,
        height_shift_range=0.2,
        rescale=1./255,
        shear_range=0.2,
        zoom_range=0.2,
        horizontal_flip=True,
        fill_mode='nearest',
        validation_split=0.25)

train_generator = train_gen.flow_from_directory(
       '/content/17flowers/jpg',  
        target_size=(224, 224),  
        batch_size=32,
        class_mode='categorical',
        subset="training")  

validation_generator = train_gen.flow_from_directory(
       '/content/17flowers/jpg',     
        target_size=(224, 224),  
        batch_size=32,
        class_mode='categorical',
        subset="validation") 

Found 1020 images belonging to 17 classes.
Found 340 images belonging to 17 classes.
In [87]:
trainXCNN,trainYCNN= next(train_generator)
testXCCNN,testYCNN =next(validation_generator)
In [15]:
checkpointFlower = tensorflow.keras.callbacks.ModelCheckpoint('/content/drive/MyDrive/GL Assignment/Convolutional Neural network/FlowersClassifierCNN.h5', #where to save the model
                                                    save_best_only=True, 
                                                    monitor='val_acc', 
                                                    mode='max', 
                                                    verbose = 1)
In [19]:
#CNN for training
#CNN for training
modelCNN = Sequential()
modelCNN.add(Conv2D(filters=128, kernel_size=3, activation="relu", input_shape=(224,224,3),kernel_initializer='he_normal'))
modelCNN.add(MaxPooling2D(pool_size=(2,2)))
modelCNN.add(BatchNormalization())
modelCNN.add(Conv2D(filters=128, kernel_size=3, activation="relu",kernel_initializer='he_normal'))
modelCNN.add(MaxPooling2D(pool_size=(2,2)))
modelCNN.add(BatchNormalization())
modelCNN.add(Conv2D(filters=64, kernel_size=3, activation="relu",kernel_initializer='he_normal'))
modelCNN.add(MaxPooling2D(pool_size=(2,2)))
modelCNN.add(Conv2D(filters=64, kernel_size=3, activation="relu",kernel_initializer='he_normal'))
modelCNN.add(MaxPooling2D(pool_size=(2,2)))
modelCNN.add(Dropout(rate=0.4))
modelCNN.add(Flatten())
modelCNN.add(Dense(256, activation="relu"))
modelCNN.add(BatchNormalization())
modelCNN.add(Dropout(rate=0.4))
modelCNN.add(Dense(128, activation="relu"))
modelCNN.add(BatchNormalization())
modelCNN.add(Dropout(rate=0.4))
modelCNN.add(Dense(64, activation="relu"))
modelCNN.add(BatchNormalization())
modelCNN.add(Dropout(rate=0.3))
modelCNN.add(Dense(17, activation="softmax"))
# Compile the model
#sgd = SGD(learning_rate=0.00027483485594610876,decay= 1e-5, momentum=0.9)
modelCNN.load_weights('/content/drive/MyDrive/GL Assignment/Convolutional Neural network/FlowersClassifierCNN.h5')
modelCNN.compile(loss="categorical_crossentropy", metrics=["accuracy"], optimizer='adam')

# Fit the model
#modelCNN.fit( x=trainXCNN, y=trainYCNN, batch_size=200, epochs=10)

historyFlowerCNN = modelCNN.fit(train_generator, 
             steps_per_epoch= 1020 // 16,
             validation_data = validation_generator,
             validation_steps = 340//16,
            callbacks = [checkpointFlower],
 epochs=10,verbose=1)

Epoch 1/10
63/63 [==============================] - ETA: 0s - batch: 31.0000 - size: 31.8730 - loss: 0.4303 - acc: 0.8606 
/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/training.py:2325: UserWarning: `Model.state_updates` will be removed in a future version. This property should not be used in TensorFlow 2.0, as `updates` are applied automatically.
  warnings.warn('`Model.state_updates` will be removed in a future version. '

Epoch 00001: val_acc did not improve from 0.77424
63/63 [==============================] - 765s 12s/step - batch: 31.0000 - size: 31.8730 - loss: 0.4303 - acc: 0.8606 - val_loss: 0.8800 - val_acc: 0.7530
Epoch 2/10
63/63 [==============================] - ETA: 0s - batch: 31.0000 - size: 31.8730 - loss: 0.4633 - acc: 0.8616 
Epoch 00002: val_acc did not improve from 0.77424
63/63 [==============================] - 771s 12s/step - batch: 31.0000 - size: 31.8730 - loss: 0.4633 - acc: 0.8616 - val_loss: 0.7438 - val_acc: 0.7727
Epoch 3/10
63/63 [==============================] - ETA: 0s - batch: 31.0000 - size: 31.8730 - loss: 0.4253 - acc: 0.8735 
Epoch 00003: val_acc did not improve from 0.77424
63/63 [==============================] - 760s 12s/step - batch: 31.0000 - size: 31.8730 - loss: 0.4253 - acc: 0.8735 - val_loss: 1.0269 - val_acc: 0.7379
Epoch 4/10
63/63 [==============================] - ETA: 0s - batch: 31.0000 - size: 31.8730 - loss: 0.4281 - acc: 0.8665 
Epoch 00004: val_acc did not improve from 0.77424
63/63 [==============================] - 765s 12s/step - batch: 31.0000 - size: 31.8730 - loss: 0.4281 - acc: 0.8665 - val_loss: 0.7732 - val_acc: 0.7500
Epoch 5/10
63/63 [==============================] - ETA: 0s - batch: 31.0000 - size: 31.8730 - loss: 0.4449 - acc: 0.8601 
Epoch 00005: val_acc did not improve from 0.77424
63/63 [==============================] - 762s 12s/step - batch: 31.0000 - size: 31.8730 - loss: 0.4449 - acc: 0.8601 - val_loss: 0.9220 - val_acc: 0.7227
Epoch 6/10
63/63 [==============================] - ETA: 0s - batch: 31.0000 - size: 31.8730 - loss: 0.4043 - acc: 0.8715 
Epoch 00006: val_acc did not improve from 0.77424
63/63 [==============================] - 778s 12s/step - batch: 31.0000 - size: 31.8730 - loss: 0.4043 - acc: 0.8715 - val_loss: 0.8774 - val_acc: 0.7318
Epoch 7/10
63/63 [==============================] - ETA: 0s - batch: 31.0000 - size: 31.8730 - loss: 0.4128 - acc: 0.8680 
Epoch 00007: val_acc did not improve from 0.77424
63/63 [==============================] - 761s 12s/step - batch: 31.0000 - size: 31.8730 - loss: 0.4128 - acc: 0.8680 - val_loss: 0.8272 - val_acc: 0.7500
Epoch 8/10
63/63 [==============================] - ETA: 0s - batch: 31.0000 - size: 31.9365 - loss: 0.3671 - acc: 0.8842 
Epoch 00008: val_acc did not improve from 0.77424
63/63 [==============================] - 777s 12s/step - batch: 31.0000 - size: 31.9365 - loss: 0.3671 - acc: 0.8842 - val_loss: 0.8714 - val_acc: 0.7394
Epoch 9/10
63/63 [==============================] - ETA: 0s - batch: 31.0000 - size: 31.8095 - loss: 0.3747 - acc: 0.8802 
Epoch 00009: val_acc did not improve from 0.77424
63/63 [==============================] - 759s 12s/step - batch: 31.0000 - size: 31.8095 - loss: 0.3747 - acc: 0.8802 - val_loss: 0.8816 - val_acc: 0.7606
Epoch 10/10
63/63 [==============================] - ETA: 0s - batch: 31.0000 - size: 31.9365 - loss: 0.3766 - acc: 0.8733 
Epoch 00010: val_acc did not improve from 0.77424
63/63 [==============================] - 766s 12s/step - batch: 31.0000 - size: 31.9365 - loss: 0.3766 - acc: 0.8733 - val_loss: 0.9189 - val_acc: 0.7424
In [20]:
modelCNN.evaluate(testXCCNN,testYCNN)
Out[20]:
[1.1702051162719727, 0.75]
In [21]:
modelCNN.evaluate(trainXCNN,trainYCNN)
Out[21]:
[0.34690654277801514, 0.875]
In [22]:
modelCNN.save("/content/drive/MyDrive/GL Assignment/Convolutional Neural network/FlowersClassifierCNN.h5")
In [23]:
predictFlowerCNN = np.argmax(modelCNN.predict(testXCCNN),axis=-1)

/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/training.py:2325: UserWarning: `Model.state_updates` will be removed in a future version. This property should not be used in TensorFlow 2.0, as `updates` are applied automatically.
  warnings.warn('`Model.state_updates` will be removed in a future version. '
In [ ]:
for i in range(20,30):
  print("Predicted value: ",predictFlowerCNN[i])
  print("Actual value: ",testYCNN[i])
  plt.imshow(testXCCNN[i])
  plt.show()

Predicted value:  1
Actual value:  [0. 1. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]

Predicted value:  11
Actual value:  [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 0.]

Predicted value:  1
Actual value:  [0. 1. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]

Predicted value:  3
Actual value:  [0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]

Predicted value:  0
Actual value:  [0. 0. 0. 0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 0. 0. 0. 0.]

Predicted value:  4
Actual value:  [0. 0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]

Predicted value:  4
Actual value:  [0. 0. 0. 0. 1. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]

Predicted value:  11
Actual value:  [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 0.]

Predicted value:  11
Actual value:  [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 0.]

Predicted value:  1
Actual value:  [0. 1. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]

4.4 CNN with transfer learning

In [24]:
#Initialise and download VGG19 model weights excluding the fully connected layer.
vgg_model = VGG19(include_top = False, weights = 'imagenet', input_shape = (224,224,3))

Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/vgg19/vgg19_weights_tf_dim_ordering_tf_kernels_notop.h5
80142336/80134624 [==============================] - 1s 0us/step
In [25]:
%cd /content/drive/MyDrive/GL Assignment/Convolutional Neural network

/content/drive/MyDrive/GL Assignment/Convolutional Neural network
In [26]:
checkpointFlowerVGG = tensorflow.keras.callbacks.ModelCheckpoint('/content/drive/MyDrive/GL Assignment/Convolutional Neural network/FlowersClassifierVGG.h5', #where to save the model
                                                    save_best_only=True, 
                                                    monitor='val_acc', 
                                                    mode='max', 
                                                    verbose = 1)
In [28]:
modelVGG= Sequential()
vgg_model.trainable = False
for layer in vgg_model.layers:
  modelVGG.add(layer)
modelVGG.add(Flatten())
modelVGG.add(BatchNormalization()) 
modelVGG.add(Dropout(0.2))
modelVGG.add(Dense(128, activation='relu',kernel_initializer='he_uniform'))
modelVGG.add(Dropout(0.2))
modelVGG.add(Dense(128, activation='relu',kernel_initializer='he_uniform'))
modelVGG.add(Dropout(0.2))
modelVGG.add(BatchNormalization()) 
modelVGG.add(Dense(40, activation='relu',kernel_initializer='he_uniform'))
modelVGG.add(Dense(17, activation='softmax'))
#learn_rate=0.001
#sgd=SGD(lr=learn_rate,momentum=.9)
modelVGG.load_weights('/content/drive/MyDrive/GL Assignment/Convolutional Neural network/FlowersClassifierVGG.h5')
modelVGG.compile(optimizer='adam',loss='categorical_crossentropy',metrics=['accuracy'])
historyVGG= modelVGG.fit(train_generator,epochs=20,validation_data=(validation_generator),callbacks = [checkpointFlowerVGG],verbose=1)

Epoch 1/10
32/32 [==============================] - ETA: 0s - batch: 15.5000 - size: 31.8750 - loss: 0.5298 - acc: 0.8431 
/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/training.py:2325: UserWarning: `Model.state_updates` will be removed in a future version. This property should not be used in TensorFlow 2.0, as `updates` are applied automatically.
  warnings.warn('`Model.state_updates` will be removed in a future version. '

Epoch 00001: val_acc improved from -inf to 0.78529, saving model to /content/drive/MyDrive/GL Assignment/Convolutional Neural network/FlowersClassifierVGG.h5
32/32 [==============================] - 885s 28s/step - batch: 15.5000 - size: 31.8750 - loss: 0.5298 - acc: 0.8431 - val_loss: 0.8502 - val_acc: 0.7853
Epoch 2/10
32/32 [==============================] - ETA: 0s - batch: 15.5000 - size: 31.8750 - loss: 0.4911 - acc: 0.8451 
Epoch 00002: val_acc improved from 0.78529 to 0.78824, saving model to /content/drive/MyDrive/GL Assignment/Convolutional Neural network/FlowersClassifierVGG.h5
32/32 [==============================] - 884s 28s/step - batch: 15.5000 - size: 31.8750 - loss: 0.4911 - acc: 0.8451 - val_loss: 0.7728 - val_acc: 0.7882
Epoch 3/10
32/32 [==============================] - ETA: 0s - batch: 15.5000 - size: 31.8750 - loss: 0.4633 - acc: 0.8471 
Epoch 00003: val_acc did not improve from 0.78824
32/32 [==============================] - 876s 28s/step - batch: 15.5000 - size: 31.8750 - loss: 0.4633 - acc: 0.8471 - val_loss: 0.7131 - val_acc: 0.7647
Epoch 4/10
32/32 [==============================] - ETA: 0s - batch: 15.5000 - size: 31.8750 - loss: 0.4633 - acc: 0.8598 
Epoch 00004: val_acc did not improve from 0.78824
32/32 [==============================] - 874s 28s/step - batch: 15.5000 - size: 31.8750 - loss: 0.4633 - acc: 0.8598 - val_loss: 0.7375 - val_acc: 0.7882
Epoch 5/10
32/32 [==============================] - ETA: 0s - batch: 15.5000 - size: 31.8750 - loss: 0.4265 - acc: 0.8578 
Epoch 00005: val_acc improved from 0.78824 to 0.83824, saving model to /content/drive/MyDrive/GL Assignment/Convolutional Neural network/FlowersClassifierVGG.h5
32/32 [==============================] - 875s 28s/step - batch: 15.5000 - size: 31.8750 - loss: 0.4265 - acc: 0.8578 - val_loss: 0.6486 - val_acc: 0.8382
Epoch 6/10
32/32 [==============================] - ETA: 0s - batch: 15.5000 - size: 31.8750 - loss: 0.4567 - acc: 0.8471 
Epoch 00006: val_acc did not improve from 0.83824
32/32 [==============================] - 876s 28s/step - batch: 15.5000 - size: 31.8750 - loss: 0.4567 - acc: 0.8471 - val_loss: 0.6762 - val_acc: 0.8118
Epoch 7/10
32/32 [==============================] - ETA: 0s - batch: 15.5000 - size: 31.8750 - loss: 0.3761 - acc: 0.8882 
Epoch 00007: val_acc did not improve from 0.83824
32/32 [==============================] - 878s 28s/step - batch: 15.5000 - size: 31.8750 - loss: 0.3761 - acc: 0.8882 - val_loss: 0.6423 - val_acc: 0.8235
Epoch 8/10
32/32 [==============================] - ETA: 0s - batch: 15.5000 - size: 31.8750 - loss: 0.3937 - acc: 0.8686 
Epoch 00008: val_acc did not improve from 0.83824
32/32 [==============================] - 874s 28s/step - batch: 15.5000 - size: 31.8750 - loss: 0.3937 - acc: 0.8686 - val_loss: 0.6448 - val_acc: 0.8324
Epoch 9/10
32/32 [==============================] - ETA: 0s - batch: 15.5000 - size: 31.8750 - loss: 0.4419 - acc: 0.8608 
Epoch 00009: val_acc did not improve from 0.83824
32/32 [==============================] - 872s 27s/step - batch: 15.5000 - size: 31.8750 - loss: 0.4419 - acc: 0.8608 - val_loss: 0.7137 - val_acc: 0.8059
Epoch 10/10
32/32 [==============================] - ETA: 0s - batch: 15.5000 - size: 31.8750 - loss: 0.4349 - acc: 0.8510 
Epoch 00010: val_acc did not improve from 0.83824
32/32 [==============================] - 874s 28s/step - batch: 15.5000 - size: 31.8750 - loss: 0.4349 - acc: 0.8510 - val_loss: 0.6097 - val_acc: 0.8147
In [29]:
#save the model for future use
modelVGG.save('/content/drive/MyDrive/GL Assignment/Convolutional Neural network/FlowersClassifierVGG.h5')
In [30]:
modelVGG.evaluate(trainXCNN,trainYCNN)
Out[30]:
[0.16627603769302368, 0.96875]
In [31]:
modelVGG.evaluate(testXCCNN,testYCNN)
Out[31]:
[0.6206525564193726, 0.8125]
In [83]:
modelVGG = load_model('/content/drive/MyDrive/GL Assignment/Convolutional Neural network/FlowersClassifierVGG.h5')

WARNING:tensorflow:From /usr/local/lib/python3.7/dist-packages/tensorflow/python/ops/init_ops.py:97: calling GlorotUniform.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.
Instructions for updating:
Call initializer instance with the dtype argument instead of passing it to the constructor
WARNING:tensorflow:From /usr/local/lib/python3.7/dist-packages/tensorflow/python/ops/init_ops.py:97: calling Zeros.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.
Instructions for updating:
Call initializer instance with the dtype argument instead of passing it to the constructor
WARNING:tensorflow:From /usr/local/lib/python3.7/dist-packages/tensorflow/python/ops/init_ops.py:97: calling Ones.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.
Instructions for updating:
Call initializer instance with the dtype argument instead of passing it to the constructor
In [88]:
#Predictions
flowerPredictions = np.argmax(modelVGG.predict(testXCCNN), axis=-1)

/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/training.py:2325: UserWarning: `Model.state_updates` will be removed in a future version. This property should not be used in TensorFlow 2.0, as `updates` are applied automatically.
  warnings.warn('`Model.state_updates` will be removed in a future version. '
In [89]:
#Compare actual values and predicted values along with the images 
for i in range(23,29):
  print("Predicted value: ",flowerPredictions[i])
  print("Actual value: ",testYCNN[i])
  plt.imshow(testXCCNN[i])
  plt.show()

Predicted value:  9
Actual value:  [0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 0. 0. 0.]

Predicted value:  5
Actual value:  [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 0.]

Predicted value:  5
Actual value:  [0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]

Predicted value:  15
Actual value:  [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 0.]

Predicted value:  3
Actual value:  [0. 0. 0. 1. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]

Predicted value:  1
Actual value:  [0. 1. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
In [ ]:
#5 out 6 of our predictions are correct

Compare the results from the above step along with your detailed observations

Accuracies for the models are as follows:

  1. Supervised learning using SVM : 50.7
  2. Neural Network : 53.8
  3. Convolutional Neural Network: 75
  4. CNN with transferred learning:81.25

Out of the four models, CNN model has performed the best.We have used VGG model for transferred learning so that the trained weights can be utlilised to improve accuracy in lesser time. Hence transfer learning with VGG19 has given an accuracy of 81.25 for our oxford flower dataset having 17 categories.

5. Create a GUI to import the image “Prediction.jpg” and use the above designed AIML image classification model to predict the class/label of this image.

In [75]:
tkWind =tk.Tk()
tkWind.title('CLASSIFIER GUI-Great Learning')
#canvas = Canvas(tkWind, width = 300, height = 300)  
#canvas.pack()  


#Step 1 :File Name
fileNamelbl = ttk.Label(tkWind,text='Step 1: File Name')
fileNamelbl.grid(row=0, column=0,sticky=tk.W)
name_var=tk.StringVar()
name_entry =ttk.Entry(tkWind,textvariable=name_var)
name_entry.grid(row=0, column=1)
#df = pd.DataFrame()
def import_data():
    base_path ='/content/drive/MyDrive/GL Assignment/Convolutional Neural network/'
    fileName =(name_var.get())
    path =base_path+fileName#
    flowerImg = cv2.imread(fileName) 
    
    #cv2_imshow(flowerImg)
    return flowerImg
# Output img with window name as 'image'
#cv2_imshow(img)

#Imprt data button
importButton =ttk.Button(tkWind,text='Import data', command=import_data)
importButton.grid(row=0,column=2)


def predict_class():
    model2 = load_model('/content/drive/MyDrive/GL Assignment/Convolutional Neural network/FlowersClassifierVGG.h5')
    #model2= load_model('FlowersClassifierVGG.h5')
    flowerImage= import_data()
    
    
    resized = cv2.resize(flowerImage, (224,224), interpolation=cv2.INTER_AREA)  
    flowerImage = np.reshape(resized,[1,224,224,3])
    predictFlower =model2.predict_classes(flowerImage)

    flowerLabels = (train_generator.class_indices)
    flowerLabels = dict((v,k) for k,v in flowerLabels.items())
    prediction = flowerLabels[predictFlower[0]] 


    predictString_lbl = ttk.Entry(tkWind)
    predictString_lbl.grid(row=1,column=3)
    predictString_lbl.insert(1,str(prediction))
    print(predictFlower)
    
predictImage =ttk.Button(tkWind,text='Predict', command=predict_class)
predictImage.grid(row=1,column=2)







tkWind.mainloop()

WARNING:tensorflow:From C:\Users\AshwiniShivaprasad\anaconda\lib\site-packages\tensorflow\python\ops\init_ops.py:97: calling GlorotUniform.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.
Instructions for updating:
Call initializer instance with the dtype argument instead of passing it to the constructor
WARNING:tensorflow:From C:\Users\AshwiniShivaprasad\anaconda\lib\site-packages\tensorflow\python\ops\init_ops.py:97: calling Zeros.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.
Instructions for updating:
Call initializer instance with the dtype argument instead of passing it to the constructor
WARNING:tensorflow:From C:\Users\AshwiniShivaprasad\anaconda\lib\site-packages\tensorflow\python\ops\init_ops.py:97: calling Ones.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.
Instructions for updating:
Call initializer instance with the dtype argument instead of passing it to the constructor

C:\Users\AshwiniShivaprasad\anaconda\lib\site-packages\tensorflow\python\keras\engine\sequential.py:450: UserWarning: `model.predict_classes()` is deprecated and will be removed after 2021-01-01. Please use instead:* `np.argmax(model.predict(x), axis=-1)`,   if your model does multi-class classification   (e.g. if it uses a `softmax` last-layer activation).* `(model.predict(x) > 0.5).astype("int32")`,   if your model does binary classification   (e.g. if it uses a `sigmoid` last-layer activation).
C:\Users\AshwiniShivaprasad\anaconda\lib\site-packages\tensorflow\python\keras\engine\training.py:2325: UserWarning: `Model.state_updates` will be removed in a future version. This property should not be used in TensorFlow 2.0, as `updates` are applied automatically.
  if class_weight:

{0: '0', 1: '1', 2: '10', 3: '11', 4: '12', 5: '13', 6: '14', 7: '15', 8: '16', 9: '2', 10: '3', 11: '4', 12: '5', 13: '6', 14: '7', 15: '8', 16: '9'}
10
[2]

PART 5: Explain in depth your strategy to maintain and support the AIML image classifier after it in production. Import image from your machine Predict image class using a trained AIML image classifier.

Some points one can follow after the model is deployed to production environment:

  1. Nature of data keeps changing all the time. So, it needs to be continuously monitored or gathered in real-time. Incoming data and distribution can be verified it is like the distribution of training data with which the model was trained. Data Changes can be tracked by few statistics like mean, variance, standard deviation, correlation etc.

  2. To validate the model’s performance in real-time, labels/predictions can be collected for a specific timeframe. Performance metrics can be calculated. This can be checked with metrics value obtained during training phase.

  3. If the above two aspects are significantly varying from original setup, then our model can be retrained with chunks of latest incoming data. Model performance needs to be continuously tracked and can be retrained accordingly as per business requirements.